CN114416493A - A kind of vehicle software running abnormal detection method - Google Patents

Abstract

The invention discloses a vehicle software operation abnormity detection method, which mainly comprises the following three steps: determining input and detection indexes of a software module, establishing a prediction model for the input and detection indexes through an AI training model, and comparing the detection indexes in the software operation process; according to the input data of the perception fusion program CPU and the CPU occupancy rate of the perception fusion corresponding moment, the relation between the CPU occupancy rate and the input data of the perception fusion program is analyzed, the CPU occupancy rate is finally predicted by using the input data of the perception fusion program, and if the deviation between the actual operation output of the program and the result of the model prediction output is overlarge, the operation is judged to be abnormal.

Description

A kind of vehicle software running abnormal detection method

technical field

The invention belongs to the technical field of detecting the running conditions of automobile software or software modules, in particular to a method for detecting abnormal running of vehicle software.

Background technique

In the field of automatic driving of vehicles, it is very important to monitor the operation of software or software modules, such as cpu occupancy rate, stack usage rate, key output variables and other indicators in real time, and determine whether it is in abnormal operation. The abnormal driving of the vehicle caused by the abnormal operation greatly improves the safety of the vehicle operation.

In the prior art, such as the patent application titled "Monitoring Method of Vehicle Domain Controller Operating Status" disclosed in CN113176771A, the disclosed monitoring method includes monitoring the occurrence of relevant important events and abnormal events, and comparing the occurrence time of the event and the event information. Cached in memory in the form of a log. By using the microcontroller unit and the microprocessor unit as the core chips of the domain controller, the microcontroller unit is used to process software functions that have predetermined requirements for real-time performance and/or security performance, and the / or software functions with predetermined requirements for communication performance, based on the reasonable allocation and redundant backup of the microcontroller unit and the microprocessor unit, so that the vehicle domain controller can take into account real-time performance, safety performance, computing performance and communication performance to meet the needs of vehicles The performance requirements of different software functions in network management effectively improve the experience of the vehicle.

However, the above descriptions mostly monitor and record the overall operation of the domain controller, not to detect a single software or software module, and it is impossible to judge whether the function of a software or software module is running abnormally. An abnormal event log management, and the definition of abnormal events is relatively vague; it cannot judge whether a certain software or software module is in an abnormal operation state at a certain time through the analysis method of big data, and it cannot prevent the occurrence of software or software modules. The abnormal driving of the vehicle caused by the abnormal operation cannot be found or eliminated in time due to the abnormal operation of the software.

SUMMARY OF THE INVENTION

In view of the above-mentioned deficiencies in the prior art, the purpose of the present invention is to provide a method for detecting abnormal operation of vehicle software, so as to judge whether a certain software or software module is in an abnormal operation state at a certain moment through the analysis method of big data, and find out in time The abnormal running of the vehicle caused by the abnormal operation of the software or software module can detect or eliminate the security risks existing due to the abnormal operation of the software in time.

The technical scheme of the present invention is realized as follows:

A method for detecting abnormal operation of vehicle software, which is characterized by comprising the following steps:

1) From the blf file, parse out the input data and CPU usage of the perception fusion program;

2) Screen the input data of the perceptual fusion program, and find out the input values of all the perceptual fusion programs corresponding to the CPU occupancy output moment;

3) Through the method of classification, predict the input data correspondence of the perceptual fusion program in all cases

The CPU usage of each item;

4) Take the input data of the perceptual fusion program as the feature variable, first standardize the feature variable;

5) Regularize CPU input items to remove items that have less impact on CPU usage;

6) Using the CPU input items as feature variables, convert these feature variables into two dimensions, convert them into pictures, input the data in the picture format into the scene prediction model MobileNet, and train the model. When the model finally begins to converge, it is regarded as a training session. Complete; continuously input data in image format into the model, optimize the model, until the final model classification accuracy reaches the set index N, and terminate the model training;

7) In the model verification stage, the program and the model run at the same time, and the input of the perception fusion program is consistent with the input of the scene prediction model. If the difference between the cpu occupancy rate sent by the scene prediction model and the actual controller is higher than the set value M, it is abnormal. . In this way, the present invention judges whether a certain software or software module is in an abnormal operating state at a certain moment through the analysis method of big data, timely finds the abnormal running of the vehicle caused by the abnormal operation of the software or software module, and then can timely find or eliminate the abnormal operation of the software or software module. The security risk of abnormal operation.

Further: in the classification method described in step 3), the parameter d is set as the interval of the CPU occupancy rate, the CPU occupancy rate is manually divided into 100/d categories, and the input items of the perceptual fusion program are used as features, and the corresponding output input The data is classified into the corresponding CPU occupancy categories, and the parameter d is gradually reduced according to the classification effect until the parameter d is reduced to 1, and the input items of the perceptual fusion program corresponding to the output items can be classified to every 1 by the classification method. % of the CPU occupancy rate, that is, each CPU occupancy rate corresponding to the input data of the perceptual fusion program in all cases can be predicted, and the CPU occupancy rate is predicted by the method of classification. In this way, since the CPU occupancy rate of the present invention and the program input items are indirectly related, the general regression method cannot achieve good results, and the regression problem needs to be converted into a classification problem. rate forecast.

Further: the standardization of feature variables described in step 4) is standard feature scaling or maximum and minimum feature scaling.

Further: the N described in step 6) is 90%. This data N is set according to the specific situation, which is convenient for adjustment according to different models.

Further: the M described in step 7) is 10%. This data M is set according to the specific situation, which is convenient for adjustment according to different models.

Further: save or output and display the abnormal result to the vehicle end. It is convenient for subsequent playback or monitoring.

Further: the scene prediction model adopts MobileNET lightweight classification network.

Further: the input data source of the perception fusion program is sensor data, including at least target data and lane line data.

In a word, the present invention has the following beneficial effects:

1. Judging whether a certain software or software module is in an abnormal operation state at a certain time through the analysis method of big data, timely discovering the abnormal operation of the vehicle caused by the abnormal operation of the software or software module, and then timely discovering or eliminating the abnormal operation of the software. security risks exist.

2. The present invention can guide developers to locate the cause of the problem, and can play functions such as warning of abnormal operation of the software.

Description of drawings

Fig. 1 is the processing flow chart of this embodiment;

Fig. 2 is the flow chart of the prediction model system of the present embodiment;

FIG. 3 is a flow chart of verification in this embodiment.

Detailed ways

The specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

Abnormalities described in the present invention: The present invention defines anomalies by means of scenario simulation, rather than simply giving a standard, and it is judged as abnormal if it exceeds or falls below a certain standard. A software function module (here we describe the perception module running on the MPU in the L3 project) for different inputs are different scenarios, different scenarios correspond to different indicators, the indicator may be the input or the cpu in the scenario Occupancy rate or the length, curvature and other characteristics of a certain lane line; if the detection index deviation is too large in similar scenarios, it means that the software is running abnormally.

Prediction model: The prediction model adopts the MobileNET lightweight classification network. The present invention takes the L3 perception fusion program as an example. The input source of the perception fusion program is sensor data, including at least target data and lane line data; the detection target is the perception fusion. The CPU usage of the program on the mpu side of the controller. The present invention selects the parameter of cpu occupancy rate as the detection index.

Referring to Figures 1-3, a method for detecting abnormal operation of vehicle software according to the present invention, the steps are as follows:

1. From the blf file (blf is the abbreviation of binary logging format, that is, binary data file, which is a format in which data packets are stored by the on-board data acquisition device. The blf file parsed here contains the on-board controller runtime. The relevant data), that is, the can message and the Ethernet intermediate variables, parse out the input data and detection indicators of the perception fusion program, that is, the cpu occupancy rate;

2. Screen the input data of the perceptual fusion program, and find out the input values of all the perceptual fusion programs corresponding to the CPU occupancy output moment;

3. Through the classification method, predict the CPU occupancy rate of each item corresponding to the input data of the perceptual fusion program in all cases; in the traditional regression prediction problem, the interpreted amount is generally regarded as a predictable value, but in this project, The correlation between CPU occupancy and program input items is not direct, or indirect correlation, therefore, the general regression method cannot achieve good results. In order to solve the problem, it is necessary to convert the regression problem into a classification problem, set the parameter d as the interval of CPU occupancy, manually divide the CPU occupancy into 100/d categories, and use the input items of the perceptual fusion program as features to convert the corresponding output The input data is classified into the corresponding CPU usage categories, and the parameter d is gradually reduced according to the classification effect. If the classification effect is ideal, the d parameter can be reduced to 1 in the end, then the perception fusion corresponding to the output item can be achieved by the classification method. The input items of the program are classified into every 1% of the CPU occupancy rate, that is, each CPU occupancy rate corresponding to the input data of the perceptual fusion program in all cases can be predicted, and the CPU occupancy rate is predicted by the classification method.

4. Taking the input data of the perceptual fusion program as the feature variable, first standardize the feature variable; the current general scheme is standard feature scaling or maximum and minimum feature scaling.

5. Regularize the CPU input items and remove the items that have less impact on the CPU usage; because there are many CPU input items, it is necessary to regularize the CPU input items and remove the items that have less impact on the CPU usage.

6. Using the CPU input items as feature variables, convert these feature variables into two dimensions, convert them into pictures, and input the data in the picture format into the scene prediction model (a common model in the prior art, because the input data is a scene-related feature) variable, so it is named scene prediction model) In MobileNet, the model is trained, and when the model finally begins to converge, it is regarded as a training completion. Continuously input data into the model and optimize the model until the final model classification accuracy reaches the set index N, such as 90% or above, and terminate the model training.

7. In the verification stage of the scene prediction model, the program and the model run at the same time, and the input of the perception fusion program is consistent with the input of the scene prediction model. If it is 10%, 12%, etc., it is judged to be abnormal. Save or output the result of the abnormal judgment to the vehicle end, and provide it to the developer to locate the problem.

A method for detecting abnormality in software operation provided by the present invention is a method for detecting abnormality in software based on big data; it is summarized into three main steps: determining the input and detection indicators of software modules, and establishing predictions from the input and detection indicators through an AI training model Comparison of detection indicators during model and software operation.

The present invention analyzes the relationship between the CPU occupancy rate and the input data of the perceptual fusion program according to the input data of the perceptual fusion program CPU and the CPU occupancy rate at the corresponding moment of the perceptual fusion program, and finally uses the input data of the perceptual fusion program to its CPU occupancy rate Prediction, if the deviation between the actual operation output of the program and the predicted output of the model is too large, it can be judged that the operation is abnormal.

Finally, it should be noted that the above-mentioned examples of the present invention are merely examples for illustrating the present invention, and are not intended to limit the embodiments of the present invention. Although the applicant has described the present invention in detail with reference to the preferred embodiments, for those skilled in the art, other changes and modifications in various forms can be made on the basis of the above description. Not all implementations can be exhaustive here. Any obvious changes or changes derived from the technical solutions of the present invention are still within the protection scope of the present invention.

Claims (9)

1. A vehicle software operation abnormity detection method is characterized by comprising the following steps:

1) analyzing input data and cpu occupancy rate of the perception fusion program from the blf file;

2) screening input data of the perception fusion programs, and finding out input values of all perception fusion programs corresponding to the CPU occupancy rate output moment;

3) predicting input data correspondence of perception fusion program under all conditions by classification method

Each item of CPU occupancy;

4) taking input data of a perception fusion program as a characteristic variable, and firstly standardizing the characteristic variable;

5) regularizing the CPU input items to remove items with small influence on the CPU occupancy rate;

6) the CPU input items are used as characteristic variables, the characteristic variables are subjected to two-dimensional transformation and are converted into pictures, data in the picture format are input into a scene prediction model MobileNet, the model is trained, and one-time training is considered to be completed when the model finally starts to converge; continuously inputting data in a picture format into the model, optimizing the model until the final model classification accuracy reaches a set index N, and terminating model training;

7) and in the model verification stage, the program and the model run simultaneously, the input of the perception fusion program is consistent with the input of the scene prediction model, and if the difference between the scene prediction model and the cpu occupancy rate sent by the actual controller is higher than a set value M, the abnormality is judged.

2. The vehicle software operation abnormality detection method according to claim 1, characterized in that: the classification method in step 3) sets the parameter d as the interval of the CPU occupancy rates, manually classifies the CPU occupancy rates into 100/d categories, classifies the input data correspondingly output into the corresponding CPU occupancy rate categories by using the input items of the perception fusion program as features, gradually reduces the parameter d according to the classification effect until the parameter d is reduced to 1, classifies the input items of the perception fusion program corresponding to the output items into each 1% of the CPU occupancy rates by the classification method, namely, each CPU occupancy rate corresponding to the input data of the perception fusion program under all conditions can be predicted, and the prediction of the CPU occupancy rates is realized by the classification method.

3. The vehicle software operation abnormality detection method according to claim 1, characterized in that: the step 4) of normalizing the characteristic variables is standard characteristic scaling or maximum and minimum characteristic scaling.

4. A vehicle software operation abnormality detection method according to any one of claims 1 to 3, characterized in that: the N in the step 6) is 90 percent.

5. A vehicle software operation abnormality detection method according to any one of claims 1 to 3, characterized in that: m in the step 7) is 10 percent.

6. The vehicle software operation abnormality detection method according to claim 4, characterized in that: and storing or outputting and displaying the result of the abnormal judgment to the vehicle end.

7. The vehicle software operation abnormality detection method according to claim 5, characterized in that: and storing or outputting and displaying the result of the abnormal judgment to the vehicle end.

8. A vehicle software operation abnormality detection method according to any one of claims 1 to 3, characterized in that: the scene prediction model adopts a MobileNET lightweight classification network.

9. The vehicle software operation abnormality detection method according to claim 8, characterized in that: the input data source of the perception fusion program is sensor data, and at least comprises target data and lane line data.

Images